Chapter 2 Data sources

We’ve found a dataset of over 720,000 competitive matches on Kaggle (https://www.kaggle.com/skihikingkevin/pubg-match-deaths), which is scraped from a game tracker website (https://pubg.op.gg/).

The dataset has two parts, one is kill_match_stats_final, which records every death that occurred within the 720k matches, the other is agg_match_stats, which contains each match’s meta information and summarizes player statistics (as provided by PUBG). The death dataset tracks each death’s weapon, killer name, killer place, killer position, victim name, victim place, victim position, and death time, etc. The aggregate dataset includes various aggregated statistics such as player kills, player damage, player distance ride, player distance walk, player team, and player team placement, etc.

We used the last two datasets of both kill and match stats, and analyzed games on the Erangel map in single player mode.

2.1 Kill data set

The kill data set contains the following variables:

kill_by: method/weapon of killing
killer_name: player name of the killer
killer_placement: killer final ranking of the game
killer_position_x: killer x-coordinate on the map, ranging from 0 to 800000
killer_position_y: killer y-coordinate on the map, ranging from 0 to 800000
map: map of the game, there are currently four maps in PUBG, which are Erangel, Miramar, Sanhok, and Vikendi
match_id: unique identification of the game, which can be mapped to the match dataset
time: death time, measured by seconds
victim_name: player name of the victim
victim_placement: victim final ranking of the game
victim_position_x: victim x-coordinate on the map, ranging from 0 to 800000
victim_position_y: victim y-coordinate on the map, ranging from 0 to 800000

2.2 Aggregated match data set

The aggregated data set contains the following variables:

date: date of the game
game_size: total amount of players in that game
match_id: unique identification of the game, which can be mapped to the kill data set
match_mode: the perspective of the game, there are two kinds of perspective in PUBG, tpp (third-person-perspective) and fpp (first-person-perspective)
party_size: amount of players in each group, there are three modes in party size, single (1 player/team), duo (up to 2 players/team), and squad (up to 4 players/team)
player_assists: player assists in each game, assist means causing damage but not death to the other players
player_dbno: play down but not out in each game, down but not out means down but got revived by teammate (only applicable in non-single mode)
player_dist_ride: distance that player traveled through vehicles
player_dist_walk: distance that player traveled by walking
player_dmg: total player damage caused on other players
player_kills: total people killed by that player
player_name: -
player_survive_time: total time passed before player dies/wins
team_id: -
team_placement: ranking of that team in that game

Erangel map. Graph from the Kaggle dataset

Note:
➔ There are several maps in PUBG including Erangel, Miramar, Sanhok, and Vikendi. Erangel is the oldest and more popular map.
➔ There are three player modes including single (1 player per team), duo (up to 2 players per team), and squad (up to 4 players per team)